What AI assistants cite when Australians ask buyer questions

We took twelve buyer questions, four in each of three sectors, and ran them through Perplexity, ChatGPT and Google AI Mode with Australian geo-targeting on 20 June 2026, then logged every source each answer cited and grouped them by type.

• Professional and B2B: choosing an SEO agency, digital marketing costs, what GEO is, B2B lead generation.
• Local services: bathroom renovation cost, finding a wedding photographer, choosing a commercial electrician, buyer's agent fees.
• E-commerce: best platform to start an online store, Shopify conversion rate, Shopify versus WooCommerce, Google Shopping.

One honest caveat: of the thirty-six answers, twenty-four returned machine-readable citations. Google AI Mode's per-answer citations did not return in this run, so the domain-level detail is drawn from Perplexity and ChatGPT. Treat the numbers as directional.

The clearest pattern was how unevenly the engines cite. Perplexity is citation-rich; ChatGPT is citation-thin. That alone changes where you start.

Roughly half of all citations went to brand or vendor sites, often the platform the question was about, with the rest spread across directories and marketplaces, editorial guides, social sources, Google itself, and a few government and industry pages.

Local services were owned by directories and aggregators. hipages and serviceseeking led trades answers, while wedding photographer answers mixed photographers' own sites with wedding directories.

E-commerce belonged to the platforms. Shopify, WooCommerce, BigCommerce and the major payment brands crowded out independent voices.

Professional and B2B was the most winnable for an agency. Answers blended a few authority hubs with named agencies, so being a clearly named, cite-able expert is achievable here.

This wider read is a separate, larger measurement than the twelve-question test above, run automatically on a different prompt set across five engines over thirty days, so read the two as complementary rather than identical. One assumption worth stating: it samples the prompts we track, not the whole market, so we treat it as a strong signal, not the last word.

The field had grown to sixty-six cited brands, yet it was no more settled. The most-cited source held just over two per cent of the share of voice, and the names were global software platforms and a few global networks, not local specialists. No Australian specialist held a meaningful position anywhere. For an Australian business prepared to become a genuinely cite-able source, that is wide-open ground.

What the assistants cite as sources barely moved. The same four domains led both months. If you want to be a source AI answers lean on, these are the neighbourhoods to earn your way into.

Which businesses get named, by contrast, was almost entirely different. Of the handful of brands the assistants named most in June, only one was still near the top a month later. The rest of the top tier turned over completely, and the wider July field ran past a hundred cited domains, a long, shifting scatter with no single name holding the answer. In a race this unsettled, no provider has locked down these questions, which is precisely the opening.

Winning AI citations takes more than a good page on your own site. Get listed and reviewed in the directories the engines trust if you are a local service, publish the definitive comparison or how-to guide if you sell online, and become a named expert if you are in professional services. And start with Perplexity, which cites the most. Tracking who gets cited and where is the grind our Snowball SEO platform handles for clients, which leaves more of the budget for the guides, video and original work the engines actually quote. For the strategy behind this, read what GEO is and how to win at it.

This is a small first pass. We will expand the sample, add a Google organic cross-reference, capture Google AI Mode citations, and re-run it quarterly so it becomes a living picture of how AI answers cite Australian businesses.